Re: [PATCH 3/9] xen/bitops: Convert 'hweight(x) > 1' to new multiple_bits_set()

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 26/08/2024 11:37 am, Jan Beulich wrote:
> On 23.08.2024 01:06, Andrew Cooper wrote:
>> Using hweight() is an especially expensive way of determining simply if
>> multiple bits are set in a value.  Worse, 4 of the 10 hweight() calls in Xen
>> are of this form.
>>
>> Switch to the new multiple_bits_set() helper.  This is far more efficient 
>> than
>> the longhand hweight() algorithm and, owing to its simplicity, likely more
>> efficient than even a dedicated instruction on a superscalar processor.
>>
>> No functional change.
>>
>> Signed-off-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
> Reviewed-by: Jan Beulich <jbeulich@xxxxxxxx>

Thanks.

>
> Just to mention it: With the title being sufficiently generic, I half
> expected the similar {bitmap,cpumask}_weight() to also be taken care of.
> Yet certainly those (and maybe also their ... == 1 forms) can be covered
> later.

They are on my radar, but are not as easy to transform.

We'd need a new bitmap API to optimise this case; looking at the
callers, it is definitely worth optimising.  There are also an awful lot
of cpumask_weight() == 1 which want converting, albeit slightly differently.

That's a full series of work in and of itself.

~Andrew